Computational complexity evaluation of ann algorithms for image steganalysis

International Journal of Latest Trends in Engineering and Technology (IJLTET)

Computational Complexity Evaluation of ANN
Algorithms for Image Steganalysis
Dr.P.Sujatha
Assistant Professor, Department of Computer Science
Vels University, Chennai, India.

Dr.S.Purushothaman
Professor, PET Engineering college, Vallioor – 627 117
Tirunelveli Dt., India,

P.Rajeswari
Research Scholar, Mother Therasa University
Kodaikanal, India,
Abstract - The major growth of information technology is based on the way how the security measures are implemented.
Steganography is a technique that implements high level security by hiding a message in a multimedia object such as
image. Steganalysis is the way of detecting such hidden messages. In order to detect the presence of hidden message,
artificial neural network algorithms such as Back Propagation and Radial Basis Function are used. This paper performs
the computational complexity evaluation of these two algorithms.
Keywords - Covert communication, Steganography, Steganalysis, ANN, Back Propagation, Radial Basis Function

I. INTRODUCTION
In today’s digital age, there are more chances for altering the information represented by an image without
leaving any traces of tampering. Many areas such as forensics investigation, surveillance systems, criminal
investigation, medical imaging, journalism and intelligence services need reliability while transferring the
information in the form of an image. Planning is the crucial part and the information planning is passed to others
through covert communication in order to hide from government and other people. The effective medium of hidden
communication is achieved by steganography. An article (Jack [11]) ensured that terrorists used steganography for
secret communication during 11th September 2001 attack.
Politicians use steganography communication to express their political thoughts that are more sensitive to
the world. The Government can take action on any politician who involves in sensitive issue like decreasing the
economical growth of the country. The ease of Internet helps in both good and bad usage. Downloading various
tools for steganography becomes a challenging task for government to trace the law breakers. The majorities of
documents used in publishing industry were digital documents with foreground (black) and background (White)
binary values. Multiple documents are manipulated everyday with binary values. Those documents are scanned and
used as a medium of steganography. Variety of data embedding algorithms and variety of images that makes the
steganography a toughest mission for researchers to develop a powerful technique for steganalysis.
II. IMAGE STEGANALYSIS
The counter-technique of image steganography is known as image steganalysis. Steganalysis begins by
identifying the artifacts that exist in the suspectable file which is a result of message embedding. The goal is not to
advocate the removal or disabling of valid hidden information such as copyrights, but to point out approaches that
are vulnerable and may be exploited to investigate illicit hidden information (Anderson et al. [1]; Johnson et al. [2];
Neil et al. [7]; Rajarathnam et al. [9]). Attacks and analysis on hidden information may take several forms like
detecting, extracting, and disabling or destroying hidden information (Westfeld et al. [3]). An attacker may also
embed counter-information over the existing hidden information. These approaches vary depending upon the
methods used to embed the information into the cover media.

Vol. 3 Issue 3 January 2014

229

ISSN: 2278-621X

International Journal of Latest Trends in Engineering and Technology (IJLTET)

A. Steganalysis Methods
Based on the way of detecting the presence of hidden message, steganalysis methods are divided as
follows.
i) Statistical steganalysis
a) Spatial domain
b) Transform domain
ii) Feature based steganalysis
Statistical Steganalysis:
Existence of the hidden message is detected using statistical analysis that is done with the pixels. It is
further classified as spatial domain steganalysis and transform domain steganalysis. In spatial domain, the pair of
pixels is considered and the difference between them is calculated. In transform domain, frequency counts of
coefficients are calculated and then histogram analysis is performed.
Feature based steganalysis:
The features of the image are used to detect hidden message in an image. They can also be used to train
classifiers.
III. RELATED WORKS
Fridrich [5] developed a steganalytic technique that detects LSB embedding in color and grayscale images.
They analyze the capacity for embedding lossless data in LSBs. Randomizing the LSBs decreases this capacity. To
examine an image, they define Regular groups (R) and Singular groups (S) of pixels depending upon some
properties. With the help of relative frequencies of these groups in the given image along with an image obtained
from the original image with LSBs flipped and with an image obtained by randomizing LSBs of the original image,
they try to predict the levels of embedding.
Fridrich [8] proposed Pairs analysis method. This approach is well suited for the embedding archetype that
randomly embeds messages in LSBs of indices to palette colors of palette image.
Westfeld [4] used visual attacks to detect the steganography by making use of the ability of human eyes to inspect
the images for the corruption caused by the embedding.
Martin [6] attempts to directly use the notion of the naturalness of images to detect hidden data. Though
they found that data hidden certainly caused shifts from the natural set, knowledge of the specific data hiding
scheme provides far better detection performance.
IV. PROPOSED ALGORITHM
Apart from all modern sciences and technologies, Artificial Neural Network (ANN) plays a vital role in
capturing and representing both linear and non-linear relationships. ANN is an intelligent system which helps to
enable machines to solve problems like human by extracting and storing the knowledge. To incorporate intelligent
method for steganalysis, this paper focuses ANN to overcome the drawbacks of the conventional steganalysis
methods. The proposed methods are,
x Back Propagation Algorithm (BPA)
x Radial Basis Function (RBF)
A. Implementation of BPA:
The BPA uses the steepest-descent method to reach a global minimum. The number of layers and number
of nodes in the hidden layers are decided. The connections between nodes are initialized with random weights. A
pattern from the training set is presented in the input layer of the network and the error at the output layer is
calculated. The error is propagated backward towards the input layer and the weights are updated. At the end of each
iteration, test patterns are presented to ANN, and the prediction performance of ANN is evaluated. Further training
of ANN is continued till the desired prediction performance is reached.

Vol. 3 Issue 3 January 2014

230

ISSN: 2278-621X

International Journal of Latest Trends in Engineering and Technology (IJLTET)

Figure 1. Detection of location of message by BPA

In Figure 1, “ƔOriginal” refers to the actual information of the image. “ Detected” information indicates
that the suspect image is a stegnographic image.
B. Implementation of RBF:
(YHU\IXQFWLRQFDQEHXQLTXHO\LGHQWLILHGE\LWVLQKHUHQWSURSHUWLHVDQGWKLVPDNHVDIRUPRIɎVXLWDEOHLQ
approximation to one problem or a particular class of problems. The selection of position and the number of centers
is similar to problems choosing the number and initial values of the weights in a multilayer perceptron (MLP). A
best approximation can be produced when optimal number of centers is identified. Neither very few nor many
centers should be chosen, since this may lead to poor approximation. It is very important to maintain equilibrium
between the number of centers and the amount of training data.

Figure 2. Detection of Message location using RBF

In the above figure, the detected information is represented by ‘’. The pixels in cover image are
represented by ‘Ɣ’.
IV. COMPARISON OF PERFORMANCES
A. Computational Complexity
The computational complexity of an algorithm is defined as the number of arithmetic operations required
for training the proposed algorithm. The performance comparison of modified ANN algorithm (Vu et al. [10]) is
studied for. Empirical formulae for computational effort are presented for BPA and RBF in Table 1.
Table 1. Computational Complexity Evaluation
Algorithm

Formula for evaluating the number of arthmetic
computation

BPA

Forward computational effort in BPA for one pattern is given by
L1

2

¦n

i 1

(n i  1)

(1.1)

i 1

Reverse computational effort in BPA for one pattern is given by
L1

2

9n L  7¦ n i n i1 
i 1

¦ (4n

TCE for BPA= {(ite) a o } n p

Vol. 3 Issue 3 January 2014

i

 5)n i1

(1.2)

i L1

231

(1.3)

ISSN: 2278-621X

International Journal of Latest Trends in Engineering and Technology (IJLTET)

TCE for RBF = {2n c 2 +inv(n c 2) n c 2 } n p (1.4)

RBF

where: TCE Total computational effort,
n c is number of centers+1(bias) ,
n p is number of training patterns,
inv is inbuilt function inverse of a matrix,
ite is the number of iteration
a o is {Forward computation in BPA + Reverse computation in
BPA}
L is the total number of layers including the input layer,
‘i’ is the layer number, and
(n i ) is the number of nodes in the ith layer

B. Computational effort comparison:
Results of the computational effort comparison of proposed algorithms are given in Table 2. By using
equations given in Table 1, the computational effort for each algorithm has been calculated and presented, in order
to compare them with regard to the total number of computational effort required by each algorithm.

6

0.000811216

111

2

RBF

2(2)

1

NA

32

MSE

Computational
effort

2(2)

iterations

Number of

BPA

Algorithm

1

S.No.

No. of nodes in
input
layer
(Hidden layer)

Table 2. Computational effort comparison for proposed algorithms.

The network trained with transformed vector requires the least computational effort. RBF algorithm needs
less computational effort than BPA.

COMPUTATIONAL
EFFORT

COMPUTATIONAL COMPLEXITY
EVALUATION
120

111

100
80
60
40
20
0

BPA
RBF
32
6
1
NUMBER OF ITERATIONS

.
V. CONCLUSION
This paper proposed the most popular supervised artificial neural network algorithms such as BPA and RBF for
detecting the presence of the hidden message. The computational effort is also compared for the proposed
algorithms. The comparison shows that RBF needs less computational effort than BPA.
REFERENCES
[1]
[2]
[3]
[4]

Anderson, R.J., and Petitcolas, F.A.P., 1998, “On the limits of steganography”, IEEE Journal on Selected Areas in Communications, Vol.
16, No. 4, pp. 474-484.
Johnson, N., and Sklansky, J., 1998, “Exploring steganography: seeing the unseen”, IEEE Computer, pp. 26-34.
Westfeld, A., and Pfitzmann, A., 1999, “Attacks on steganographic systems”, Lecture notes in computer science, proceedings of the 3rd
International Workshop on Information Hiding, Vol. 1768, pp. 61-76.
Westfeld, A., and Pfitzmann, A., 2000, “Attacks on Steganographic Systems”, Lecture Notes in Computer Science, Springer-Verlag, Berlin,
Vol. 1768, pp. 61-75.

Vol. 3 Issue 3 January 2014

232

ISSN: 2278-621X

International Journal of Latest Trends in Engineering and Technology (IJLTET)

[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

[15]

[16]
[17]
[18]

Fridrich, J., Goljan, M., and Du R., 2001, “Reliable Detection of LSB Steganography in Color and Gray-Scale Images”, Magazine of IEEE
Multimedia Special Issue on Security, pp. 22-28.
Martin, A., Sapiro G., and Seroussi G., 2004, “Is image steganography natural?”, Information Theory Research Group, HP Laboratories
Palo Alto, HPL-39(R.1).
Neil Provos, and Peter Honeyman, 2003, “Hide and Seek: An Introduction to Steganography”, IEEE Security & Privacy, Vol. 1, No. 3, pp.
32-44.
Fridrich, J., and Goljan, M., 2003, “Digital image steganography using stochastic modulation”, Proceedings of IST/SPIE’s 15th Annual
Symposium on Electronic Imaging Science and Technology, San Jose, CA.
Rajarathnam, Chandramouli, Mehdi Kharrazi, and Nasir Memon, 2004, “Image Steganography and Steganalysis: Concepts and Practice”,
Lecture Notes in Computer Science, International Workshop on Digital Watermarking, Korea, Vol. 2939, pp. 35-49.
Vu Dao, N.P., and Rao Vemuri, 2002, “A Performance Comparison of Different Back Propagation Neural Networks Methods in Computer
Network Intrusion Detection”, Differential Equations and Dynamical Systems, Vol. 10, No. 1, pp. 201-214.
Jack Kelley., 2001, “Terror Groups Hide Behind Web Encryption”, USA Today.
Natarajan, V., and Anitha, R., 2012, “Blind Image steganalysis Based on Contourlet Transform”, International journal on Cryptography &
Information Security, Vol. 2, No. 3, pp. 77-87.
Chin-Chen Chang, Lin, CY., and Fan, YH., 2008, “Lossless data hiding for color images based on block truncation coding”, Pattern
Recognition, Vol. 41, No. 7, pp. 2347–2357.
Yun Shi, Q., Guorong Xuan, Chengyun Yang, Jianjiong Gao, Zhenping Zhang, Peiqi Chai, Dekun Zou, Chunhua Chen, and Wen Chen,
2005, “Effective steganalysis Based on Statistical Moments of Wavelet Characteristic Function”, IEEE International Conference on
Information Technology: Coding and Computing, Vol. 1, pp. 768-773.
Yong Wang, JiuFen Liu, and WeiMing Zhang, 2009, “Blind JPEG steganalysis Based on Correlations of DCT Coefficients in Multidirections and Calibrations”, International Conference on Multimedia Information Networking and Security, MINES'09, Vol. 1, pp. 495499.
Xiao Yi Yu, and Aiming Wang, 2009, “Steganalysis Based on Bayesion Network and Genetic Algorithm”, 2nd International Congress on
Image and Signal Processing, CISP'09, pp. 1-4.
Declan Mc Cullah., 2001, “Secret Messages Come in Wavs”, Wired News.
Daniel Lerch-+RVWDORWDQG'DYLG0HJÕDV“Steganalytic Methods for the Detection of Histogram Shifting Data Hiding Schemes”,
Proceedings of Reunión Española Cryptology and Information Security (RECSI).

Vol. 3 Issue 3 January 2014

233

ISSN: 2278-621X

Documento similar